Does Unlearning Truly Unlearn? A Black Box Evaluation of LLM Unlearning Methods